Font Size: a A A

Nose Pore Recognition And Finger Vein Recognition

Posted on:2010-08-20Degree:DoctorType:Dissertation
Country:ChinaCandidate:S L SongFull Text:PDF
GTID:1118360302983782Subject:Radio Physics
Abstract/Summary:PDF Full Text Request
With the development of information technology,biometric recognition has gone into deep life.It is based on the inherent feature of people to authenticate and recognition.It overcomes the shortcomings of the conventional information safety technology,and offers a more convenient,safer and more stable information safety technology.It will be a reform to the information safety technology.Biometric recognition has been developed fast in last decade and is becoming an application technology time by time.Recently main researchers pay attention on the way for promoting higher correct recognition rate and looking for more stable system.For better stable and better correct recognition system,biometric features progress from rough features to minute features.Face recognition is one of mature technologies in biometric relatively.But its correct recognition rate is easy influenced by many factors,as its pose,pretend.So people work on different way to improve the human's biometric recognition correct.A new biometric recognition way-nose pore is proposed in this paper,which can even be fused with face recognition to improve its correct recognition rate.Sweat pores,found in humans and other mammals,are epidermal structures responsible for the secretion of sebum.The pore as the terminal of a sweat gland and its distribution is unique to each individual and constant throughout one's life.Pore features have been claimed to be permanent,immutable,and unique according to forensic experts,and if properly utilized,can provide discriminating information for human identification.Vein pattern is looked as one of biometric recognition pattern with the likely to develop in a desirable manner.In this paper,the capture,enhancement and extraction of finger vein pattern is studied.The main results of this study are summarized in the following:(1) A new biometric recognition way,nose pore recognition is proposed.The nose pore as the terminal of sweat gland,has a fit construct which can not change in one's life.It come into being with skin,and keep on line with skin.The skin on nose does not change after growing up.So,the construct of nose pores is more stable.Two segmentation ways were used to sengment SOI.The position of nose pore is used for recognition.The nose pore pattern is extracted by the Hessian eigenvalues through the relationship between eigenvector and gradient,and then classified using a correlated systematic tool.The taken nose pore pattern can be different when the person changes his(her) nose posture.Therefore,the nose posture was estimated through the change of tow nostril area ratio,and an adjustment is done by an affine transform based on the ratio change to reduce the influence of the nose posture on the taken patterns.In this way the nose pore recognition shows a nice stability and acceptable high recognition accuracy 88%.The nose pore pattern can be fused with face to be a double modal recognition system.(2) We propose an improved locality preserving projections algorithm,called weighted locality preserving projections(WLPP).Locality preserving projections(LPP) can optimally preserve the neighborhood structure of the data set and project the data along the directions of maximal variance.LPP deemphasizes discriminative information, but DLPP reflects a good separation between classes,which is important for recognition problem.In WLPP,the similarity matrix was weighted by the difference between the numbers of two nose pore images,which aims at preserving the local manifold structures and improving the separation information by critical feature.The experimental results based on the unique nose pores database demonstrated a higher correct recognition rate for biometric identification,93.24%.Compared with results without using WLPP,the feature extraction by WLPP is more precise for the nose pore recognition.(3) Improving of the finger vein capture device and finger vein recognition.LED is used as the light source in previous devices.For the lower penetrability and little power of LED,the captured images contain not only vein patterns but also irregular shading and noise.There are some expand and shrink in blood vessel corresponding to the high and low temperature.So,there is some change in the finger vein image under different temperature.In this paper,LD array is used as the light source.LD with high penetrability can decrease the noise of image,so the captured image will be influenced by the changing of environmental temperature hardly.A multi-resolution method is utilized to improve the S/N ratio of the finger vein images,then based on the first joint location of fingers the vein images are deal with partitioning.The feature of finger vein images is extracted using of the blanket dimension and lacunarity of fractal theory.Here the simulation results show the effective corrective recognition rate for finger vein modal is over 97%.
Keywords/Search Tags:finger vein pattern recognition, nose pore recognition, blanket dimension, lacunarity, Hessian matrix, eigenvalue, eigenvector, gradient, LPP, DLPP, WLPP, affine transform
PDF Full Text Request
Related items